Prepare Dataset (scannet)

[r03943158]

Hi, thank you provide awesome work!

When I preparing the scannet, I am curious the command:

python extract_pointcloud.py path/to/extracted_data/depth/ path/to/extracted_pointcloud_visualization/ path/to/extracted_pointclouds/ scannet_datalist.npy

The argv[1] seems that the scans/depth generated from step1 and the argv[3] is the path where put the extracted pcd.
What path should I put in the argv[2] or how to ignore the argv[2]?

Thank you!

[imisra]

@zaiweizhang might know this :)

[zaiweizhang]

argv[2] are just visualizations. You can put it anywhere you want. If you want to ignore them, the most important line to comment out is this one: https://github.com/facebookresearch/DepthContrast/blob/b8257890c94f7c58aeb5cefeb91af031692611d6/data/scannet/extract_pointcloud.py#L78

[r03943158]

@zaiweizhang @imisra Thank you provide for the precise details! I can run the second command.

However, I face another issue that when I extract some frames, the process occurs error at line 79 and reports the message from o3d such as "RPly: Failed writing y of vertex 303 (binary output: double)", "RPly: Failed writing blue of vertex 304 (binary output: uchar)"but the point cloud is not empty.

Is it caused by using python3 to run the second step or other possible issues? Could you give me some advice about this?

Thank you!

[zaiweizhang]

Did you install open3d? If not, please install it like the following: http://www.open3d.org/docs/release/getting_started.html